Volunteering/Consulting for libm Development in glibc

[Adhemerval Zanella Netto]

On 17/02/25 12:34, Ayan Shafqat wrote:
> Hello All:
> 
> I am writing to express my interest in contributing to the libm portion
> of the GNU C Library. I have over a decade of experience in maintaining
> and optimizing math libraries across multiple platforms, and I’d be glad
> to apply my expertise to optimize, maintain, and fix bugs in libm.

Hi Ayan,

Thanks for taking interest in the glibc project.  We don't have a defined
project desirable features for current version (the one in wiki is quite
old and I think not really representative of current development), but I
think there are some areas where any help would be welcome:

  * The math subsystem bug reports has some backlog [1], although most of them
    for correctness and also for some not usual architectures.  I think some
    bug triage is also welcome, I think some bug reports are outdated and
    can be closed.

  * Finish c23 support, which Joseph has started and he can fill the missing
    spots.  The glibc still lack support for some functions (compoundn, pown,
    powr, rootn, rsqrt).

  * Evaluate current implementations, including performance (latency, throughput),
    correctness, and error bounds.  Paul Zimmermann has been helping us on this,
    and although our testing are quite comprehensible there some rough spots
    (like proper document the expect error ranges and performance).  A good start
    would to improve the benchtests for math functions, by adding missing inputs
    for some symbols.

  * Improve current implementations, which is something I am focusing lately.
    I am currently evaluating the CORE-MATH project, and I am already integrated
    a lot of they implementation on glibc. There are still a lot of work to do,
    mainly to check on double and long double (only x86_64 afaik).    

[1] https://sourceware.org/bugzilla/buglist.cgi?bug_status=UNCONFIRMED&bug_status=NEW&bug_status=ASSIGNED&bug_status=SUSPENDED&bug_status=WAITING&bug_status=REOPENED&component=math&list_id=90317&product=glibc&query_format=advanced
[2] https://gitlab.inria.fr/core-math/core-math/

> 
> Some of my recent highlights include:
> 
> Ardour (ardour.org): Optimized Ardour's mixing routines for x86_64 with
> SSE, AVX, and AVX512F instruction sets, as well as ARM NEON (32-bits)
> instruction sets.
> 
> A5Eq.lv2 (github.com/ashafq/A5Eq.lv2): Optimized IIR filter for x86_64
> with AVX, SSE, and ARM NEON (64-bits).
> 
> I am happy to assist with any math-specific tasks in the C library.
> Please feel free to reach out if there are areas I can help with. Thank
> you in advance, and I look forward to collaborating with you.
> 
> Best Regards,
> Ayan